1/2
PhD Proposal at CEDRIC laboratory of CNAM School
Machine Learning for Efficient Massive MIMO processing in Beyond 5G and 6G Summary
Research areas: wireless communications, machine leaning
Director : Prof. Didier LE RUYET (CEDRIC/LAETITIA,[email protected]) Supervisors : Dr. Rafik Zayani (CEDRIC/LAETITIA, [email protected])
Dr. Marin Ferecatu (CEDRIC/VERTIGO, [email protected]) Dr. Nicolas Audebert (CEDRIC/VERTIGO, [email protected]) Start date : September 2020 or later
Laboratory : CEDRIC/CNAM, 2, rue Conté, 75003 Paris France
Context of the research
Massive multi-user MIMO (MU-MIMO) [MAR10] has been recognized as one of the most promising approaches for future generations (beyond 5G and 6G) of wireless systems, representing the most ultimate enablers of enhanced energy-efficiency (EE) and spectral-efficiency (SE) [LAR14]. However, massive MIMO deploys as many RF transceivers as there are antennas, which are expected to be inexpensive to enable cost- and energy-efficient massive MU-MIMO BS deployments. Furthermore, the associated hardware impairments require advanced signal processing, threatening the massive MIMO’s good qualities related to SE and EE. Some techniques have been proposed in order to allow the use of energy-efficient RF frontend [KON18][ZAY19][ZAY19b], but their complexity are still an open challenging issue, especially with fast time-varying channel and large number of users. Therefore, it is highly desirable to identify new approaches to reduce the complexity of the processing required, making massive MU-MIMO more attractive for B5G and 6G.
On the other hand, deep learning (DL) based machine learning (ML) has been shown to be an advanced tool capable of building universal classifiers and/or approximate general functions. Previous works using artificial neural networks (ANN) have been able to efficiently perform the real-time optimization required for massive MU-MIMO (see for example [WAN19] [WEN18] [HUA18] [DEM20]). However, these classical DL methods adopt a data-driven approach to identify the most adequate architecture of an ANN that fits well input-output data pairs where a huge amount of live data is required, which is not practical in real-time systems due to fast-varying, complexity, energy consumption in data acquisition process.
In order to overcome this problem, in the deep learning community new advanced approaches are currently being investigated in order to complement purely data-driven Machine Learning (ML), such as knowledge-driven ML [ZAP19] (i.e., using prior information based on theoretical models) and meta- learning [CHE20] (learning to learn) approaches.
Objectives
This PhD thesis will study technologies and opportunities available to embrace DL to tackle the aforementioned massive MIMO’s issues. Specifically, it will investigate the combination of the two approaches of knowledge-driven ML and meta-learning to (i) allow the use of low-cost and low-power hardware components, such as power amplifiers (PAs) and ADC/DAC (PAPR reduction and nonlinear distortion problems will be addressed); (ii) channel estimation under hardware imperfection (the main challenge will be the reduction of the training overhead); (iii) radio resource management for real-time energy-efficiency maximization (PA imperfection and channel estimation error will be considered). Our goal will be to replace or improve existing optimization algorithms by approximating them with efficient fast deep networks. Reducing the computational cost of neural networks at inference time through quantization or compression can also be investigated [CHE19].
2/2
References
[MAR10] T. L. Marzetta, "Noncooperative Cellular Wireless with Unlimited Numbers of Base Station Antennas," in IEEE Transactions on Wireless Communications, vol. 9, no. 11, pp. 3590-3600, November 2010.
[LAR14] E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta, “Massive MIMO for next generation wireless systems,” IEEE Commun. Magazine, vol. 52, no. 2, pp. 186–195, Feb. 2014.
[WAN19] T. Wang, C. Wen, S. Jin and G. Y. Li, "Deep Learning-Based CSI Feedback Approach for Time- Varying Massive MIMO Channels," in IEEE Wireless Communications Letters, vol. 8, no. 2, pp. 416-419, April 2019.
[WEN18] C. Wen, W. Shih and S. Jin, "Deep Learning for Massive MIMO CSI Feedback," in IEEE Wireless Communications Letters, vol. 7, no. 5, pp. 748-751, Oct. 2018.
[HUA18] H. Huang, J. Yang, H. Huang, Y. Song and G. Gui, "Deep Learning for Super-Resolution Channel Estimation and DOA Estimation Based Massive MIMO System," in IEEE Transactions on Vehicular Technology, vol. 67, no. 9, pp. 8549-8560, Sept. 2018.
[ZHA19] Yu Zhang and Muhammad Alrabeiah and Ahmed Alkhateeb, “Deep Learning for Massive MIMO with 1-Bit ADCs: When More Antennas Need Fewer Pilots”, arXiv 1910.06960, Oct 2019.
[ZAY19] R. Zayani, H. Shaiek and D. Roviras, "PAPR-Aware Massive MIMO-OFDM Downlink," in IEEE Access, vol. 7, pp. 25474-25484, 2019.
[ZAY19b] R. Zayani, H. Shaïek and D. Roviras, "Efficient Precoding for Massive MIMO Downlink Under PA Nonlinearities," in IEEE Communications Letters, vol. 23, no. 9, pp. 1611-1615, Sept. 2019.
[KON18] C. Kong, A. Mezghani, C. Zhong, A. L. Swindlehurst and Z. Zhang, "Multipair Massive MIMO Relaying Systems With One-Bit ADCs and DACs," in IEEE Transactions on Signal Processing, vol. 66, no. 11, pp. 2984-2997, 1 June1, 2018.
[DEM20] Ö. T. Demir and E. Björnson, "Channel Estimation in Massive MIMO Under Hardware Non- Linearities: Bayesian Methods Versus Deep Learning," in IEEE Open Journal of the Communications Society, vol. 1, pp. 109-124, 2020.
[ZAP19] A. Zappone, M. Di Renzo, M. Debbah, T. T. Lam and X. Qian, "Model-Aided Wireless Artificial Intelligence: Embedding Expert Knowledge in Deep Neural Networks for Wireless System Optimization," in IEEE Vehicular Technology Magazine, vol. 14, no. 3, pp. 60-69, Sept. 2019.
[CHE20] D. Chen, Y. Liu, B. Kim, J. Xie, C. S. Hong and Z. Han, "Edge Computing Resources Reservation in Vehicular Networks: A Meta-Learning Approach," in IEEE Transactions on Vehicular Technology.
[CHE19] Y. Cheng, D. Wang, P. Zhou, T. Zhang, “A Survey of Model Compression and Acceleration for Deep Neural Networks”, in IEEE Signal Processing Magazine.
